Short-Term Scientific Mission

Towards More General Metaheuristics for Multi-Objective Mixed Numerical-Combinatorial Optimisation Problems

Main theme: Single- and Multiobjective Optimisation
Grantee: João António Fialho Vilas Boas Duro, University of Sheffield, United Kingdom
Host: Vanessa Volz, Centrum Wiskunde & Informatica (CWI), Amsterdam, Netherlands

Start date: 2026-05-04
End date: 2026-05-15
Awarded: 2026-03-29
Report approved: 2026-07-02

Description

The aim of this STSM is to improve the generability of the design of metaheuristics for solving real-world mixed numerical-combinatorial multi-objective optimization problems. To achieve this the goals are to: identify combinatorial real-world multi-objective optimization problems, and specific problem instances, yet to be solved by metaheuristic approaches; implement and demonstrate a model-ROA interface in the Liger software that facilitates the separation of the problem model from the solver specification; and identify a collection of specialised optimization algorithm components that have potential to be ported into the model-ROA interface. The planned activities include giving a seminar at Centrum Wiskunde & Informatica (CWI).

Achievements

This STSM proposed a model-ROA interface (MRI) for multi-objective combinatorial optimization. Focusing on a novel real-world public administration problem, problem domain knowledge has been integrated directly into the variation operators of NSGA-II, and has been ported to the MRI. This work has been open-sourced in https://ligerdev.shef.ac.uk/liger-development-team/mootailor. Adapting other algorithm tailored components to the MRI remains an open issue. Future work will include studying more thoroughly the impact of algorithm tailoring, and working on a forthcoming research paper. This STSM has laid the foundations for a research collaboration focusing on the interface between algorithm tailoring and generalisability.